Why the A99 is slower than the A77 :-)

EinsteinsGhost said:

VirtualMirage said:

No, it is a budget limitation then. The technology and engineering know how is already out there...

Click to expand...

Unlimited data transfer rate? Show me, at any price.

Click to expand...

Now you are just being silly. That comment is like a kid saying "My G.I. Joe is better than yours times infinity."

We aren't talking about unlimited data transfer rate. We are talking about transfer rates that are faster than what is in the A99 or most other cameras, for that matter.

What is in today's camera is not necessarily a limitation of the technology available to them, it is a limitation of what can they afford to put in to the camera and sell it at a price point they wish to put it on the market for. Thus, it isn't an engineering limitation. It is a budget limitation.

Budgets trump engineering almost every time.

But to answer your question:

The Red Epic camera can capture 5k (13.8MP) images at up to 120fps at full resolution in both 12-bit and 16-bit RAW. And if you think it is just for cinema use, it is also used for photography. They plan to ship in the near future their Red Dragon sensor which can capture 6k (24MP) at a rate of 85fps or faster at full resolution. They also have in the pipeline a Epic 645, which is a medium format sensor that can capture 9k ( around 65MP) images at up to 50fps.

There are others out there, but the ones mentioned above are probably some of the most recognizable.

But when you start getting into frame rates that high, your limitation of maintaining that frame rate will now be limited to your lighting and equivalent shutter speed.

--
Paul

JohnFrim said:

Disagreements often arise out of misunderstanding, and perhaps we have misunderstood each other on some fine points. I tried pulling this apart point by point and reassembling it so that we might see eye to eye, but the discussion was getting far too long. A better approach might be to clarify some high level concepts; we could get into the weeds later if you like.

Let's start with your statement near the end of your post where you said:

JohnFrim said:

So assuming that only the cropped portion of the full frame needs to be covered by the shutter for a proper exposure, the speed of the full frame shutter does not need to change that much (if at all).

Click to expand...

In a follow-up post (in the original thread) you responded to altendky's suggestion (that your theory only works if the FF shutter travels a restricted distance) by saying:

JohnFrim said:

Why would you need to design the shutter to travel two different distances?

Click to expand...

JohnFrim said:

A crop mode resides within the center of a full frame and covers less area. The larger shutter distance will cover the crop mode area just as well. Since there is less distance that has to be covered during the shutters travel to black out the crop mode area, the full frame shutter's traveling speed at 6fps would be equivalent to a faster fps for the crop mode with little to no increase done to the shutter motor.

Click to expand...

I took from these statements that you believe a FF shutter traveling the full-frame 23.8 mm vertical distance during 6 fps burst operation will magically provide a faster equivalent frame rate on a crop area of 15.6 mm vertical distance with no increase in traverse speed. Read your statements again -- they really imply that a shutter operating at 6 fps can give higher (10?) fps simply by selecting crop mode on the sensor.

Is that what you were saying, or did I misinterpret?

The only way to get a higher frame rate is to increase the number of full-cycle operations of the shutter each second. Period. (Or is this just my theory that you might some day disprove? -- Sorry, I just could not resist this little dig ;-) )

Now, there are a number of ways to reduce the cycle time of the shutter that include speeding up traverse times (and re-cocking times), and reducing the traverse distance so the next cycle can start sooner, neither of which you say are necessary simply because the vertical crop area is smaller.

Respectfully,
JF

Click to expand...

I am not ignoring you here. It's just the responses I would like to type would take up more time than what I have been available to do so the past couple of days. Before posting on topics like this, I usually spend a lot of time researching and/or drafting and rewriting what I type several fold.

I would be interested in discussing this further and wonder if it would be more appropriate in a separate thread that is dedicated to the topic. That way anybody else that would like to contribute is contributing directly to the subject at hand.

I'd love to be able to learn more from everyone else's contributions and I wouldn't have any issues saying I am wrong if evidence and examples prove otherwise. Again, it is just a theory of mine and I haven't been able to find much information elsewhere to say otherwise or that my ideas are impossible (although they might not be implemented in full in today's tech if that is the case).

I've been tempted of trying to contact someone like Paul Genge via his new blog to see if he can enlighten us or knows any photography engineers that can explain the inner workings in a bit more detail. But alas, time is limited.
--
Paul

Thanks for the reply. I agree a detailed discussion would take far too much space, and that is why I started at the higher level of understanding the concept.

Summarizing my point and ignoring a bunch of other factors for the moment, the same shutter traverse speed can only contribute to a higher fps when comparing FF to APS-C cameras because the sensor size, and therefore the required traverse distance, is smaller on the APS-C. And this shorter traverse distance is important not only for how far the shutter travels in the expose phase, but also in how far the shutter has to be moved during reset. In other words, shorter operating distance with the same speed of the mechanicals will allow the shutter to complete one operating cycle in less time, and it can then be tripped sooner for the next frame.

But using a FF camera in crop mode and running the same shutter traverse speed cannot give a higher fps. The required sensor area exposure time is certainly shorter by the ratio of the vertical frame height when in crop mode, but if the shutter still travels the FF distance it will not result in a higher equivalent fps.

That was the statement I was objecting to.

The other basis of my saying your theory is weak is that your math was based on correct values regarding cycle times for 6 vs 10 fps, but you translated that into shutter traverse speeds of mm/s. If you look back at the video links you posted it is really obvious that shutter traverse duration during the expose phase is a very very small fraction of the entire shutter cycle. It just does not make sense to use these shutter traverse speeds as the basis for comparison, especially when your theory is based on a mm/s traverse speed that represents movement in one direction over the entire cycle period.

As I have stated, higher fps requires shorter total shutter cycle times. I would bet that to keep things simple the shutter mechanism is designed to operate at some maximum rate like 10 fps and it is slowed down for 6 fps. And slowed down means a delay before tripping rather than a change in actual cycle time between being tripped and being re-cocked for the next firing.

VirtualMirage said:

EinsteinsGhost said:

VirtualMirage said:

No, it is a budget limitation then. The technology and engineering know how is already out there...

Click to expand...

Unlimited data transfer rate? Show me, at any price.

Click to expand...

Now you are just being silly. That comment is like a kid saying "My G.I. Joe is better than yours times infinity."

We aren't talking about unlimited data transfer rate. We are talking about transfer rates that are faster than what is in the A99 or most other cameras, for that matter.

Click to expand...

You were talking about data transfer rate only being limited by cost. Now you're changing the tune. Is it dependent on cost, or is it dependent on available technology? Cost, of course will play a role, but let us not forget that despite the cost, technology has its limits at any given time.

BMWX5 said:

Well 1DX is worth $6799.00. For that price I am sure Sony can make a processor faster than 1DX processor!

Endos said:

http://www.youtube.com/watch?v=i3fG018tmMc

--
Regards,
Juan

http://500px.com/endosphoto
http://www.endosphoto.com

Click to expand...

Click to expand...

True, which is why the Canon 1DX has three processors. Two DIGIC 5+ being used as ISP (I guess) and one DIGIC 4 just for the AF system.

With mutiple core Bionz processing I doubt it that the A99 would be able to cost under $3K considering the rich feature list of the Sony FF. I think the A99 has one latest gen. Bionz.

EinsteinsGhost said:

You were talking about data transfer rate only being limited by cost. Now you're changing the tune. Is it dependent on cost, or is it dependent on available technology? Cost, of course will play a role, but let us not forget that despite the cost, technology has its limits at any given time.

Click to expand...

Either we are not on a same page or it is you that is changing the tune.

Let me recap what our discussion has been about and correct me if there is something wrong here:

You were claiming that the frame rate on the A99 was limited by bandwidth. I was telling you that bandwidth is not the limiting factor in the camera. You state that it is and that it appears the maximum throughput of these cameras today caps at around 300MB/s. I then tell you that is not the case and offer some examples where it exceeds that. You continue to state bandwidth is the limitation and that it is an engineering limitation because the technology is not yet there to exceed this + or - 300MB/s figure. I respond back stating that the technology is here, that there are cameras out there with a faster bandwidth capability than what you mention.

I then ask, to get better clarification, as to where in the pipeline of the camera (the A99) do you feel there is a bandwidth limitation. You state that it is the whole pipeline from beginning to end. I then tell you that if the bandwidth is limited from beginning to end then it is not limitation, it was made that way by design. You then say that it is an engineering limitation. I retort back saying that it would not be an engineering limitation if that was the case, that it would be a budget limitation. Reason being is that there are cameras out there that can push through more data from beginning to end than the A99 could ever dream about, so the technology and engineering know how is already there. It is possible. What becomes the limiting factor here is not what is possible from a technology or engineering standpoint, but a budget standpoint. Because at this point, we are no longer limited by what is possible with today's technology, we are limited by how much the company is willing to spend to implement the technology and still be able to sell it at a price point that they want to target.

You then throw in a ridiculous figure in the form of a question of unlimited transfer rate and request an example of a camera that can do this. I tell you that your request for unlimited transfer rate and an example showing such was a silly response and I try to get you back on track to what we were initially discussing, that the A99 is not limited by bandwidth per your 300MB/s number you were throwing out. To play along though, I decided to give you examples of other cameras out there that have a very high frame rate with a bandwidth throughput that would put the A99 to shame.

I never discussed or said that we can achieve unlimited bandwidth throughput, that is impossible. I was saying that we can already achieve a higher bandwidth throughput than what you claim is possible, which your claim was around the 300 MB/s mark. I was also stating that the limitation in the A99 that cap it at 6 fps I do not believe is due to a bandwidth limitation.

The limitation for the A99's frame rate is either due to mechanical reasons (shutter and/or aperture control motor) or another possible theory I am working on is processing muscle. But it isn't bandwidth.

So, am I on the right track with our discussion?

Now I haven't mentioned processing muscle much because I was still trying to iron out the details and do more research. But I'll give you my rough draft of an idea.

From an image processing only perspective, there should be no reason the BIONZ chip can't handle the A99's resolution at 14-bit at a higher frame rate than 6fps providing it is the same or even just slightly improved version over the A77. The proverbial wrench, though, that can possibly bog the processor down depends on if the processor has to share resources with other tasks or if it has dedicated resources to just do image handling. If it is just doing image handling, there should be no problems. But if it has to share resources, then it could be hampered by the addition of the new 102 PD AF points on the sensor (which the A77 didn't have) and the new AFD mode (which the A77 also didn't have). If the calculations needed to handle these tasks are being shared by the same chip pooling from the same resources and the chip was not much improved since the A77, then that COULD effect the frame rate. But again, that isn't bandwidth. That is processing power, algorithm calculating, number crunching.

What makes it tough to find out what is under the hood is:

1) We haven't seen a tear down of the A99 nor the A77 so we can't see if it is the same chip.

2) Even if we did see a tear down, we wouldn't know the clock speed that they run at.

3) Sony names all their chips in their Alpha lineup (and I think NEX lineup) since the A700 as 'BIONZ' despite the chips being very different. This makes it complicated to figure out what varies from one chip to the next other than reading Sony's marketing documentation about improvements made from last generation (they haven't mentioned much lately).

--
Paul

The hi-speed burst frame rate for the A99 (6fps) is 3/4 or 75% of the frame rate of the A77 (8fps) when comparing apples to apples, that is in full aperature / full autofocus mode.

Goetz48 wrote:
"So it is no wonder that the frame rate hase reduced to the half. "
--
Domenick Creaco

VirtualMirage said:

Let me recap what our discussion has been about and correct me if there is something wrong here:

Click to expand...

Of course. I hate to trim your posts in my responses though but here is why I do it. You start with an assumption, and work with those throughout, no matter how far off they might be from reality. Let us try to work in smaller chunks.

VirtualMirage said:

You were claiming that the frame rate on the A99 was limited by bandwidth. I was telling you that bandwidth is not the limiting factor in the camera. You state that it is and that it appears the maximum throughput of these cameras today caps at around 300MB/s. I then tell you that is not the case and offer some examples where it exceeds that.

Click to expand...

No you didn't. All you did was make assumptions that fit your whim and drew conclusions from it. If Sony A99 is producing 43-44 MB RAW files, and you claim that about 300 MB/s is non-sense, then show me a camera where you see 450-500 MB/s system. Got ONE example?

You've also assumed that in crop mode, the mechanical shutter has a mechanism to drop down to the beginning of "cropped area" and goes no farther than the other end of the cropped area. Got anything to support this claim?

EinsteinsGhost said:

VirtualMirage said:

Let me recap what our discussion has been about and correct me if there is something wrong here:

Click to expand...

Of course. I hate to trim your posts in my responses though but here is why I do it. You start with an assumption, and work with those throughout, no matter how far off they might be from reality. Let us try to work in smaller chunks.

EinsteinsGhost said:

You were claiming that the frame rate on the A99 was limited by bandwidth. I was telling you that bandwidth is not the limiting factor in the camera. You state that it is and that it appears the maximum throughput of these cameras today caps at around 300MB/s. I then tell you that is not the case and offer some examples where it exceeds that.

Click to expand...

No you didn't. All you did was make assumptions that fit your whim and drew conclusions from it. If Sony A99 is producing 43-44 MB RAW files, and you claim that about 300 MB/s is non-sense, then show me a camera where you see 450-500 MB/s system. Got ONE example?

Click to expand...

I gave you several. You just refuse to accept them. You flip my numbers and began comparing compressed RAW files to uncompressed RAW files so as to suit your 300MB/s figure. You can't compare compressed RAW to uncompressed RAW. The data off the sensor is not reaching the processing chip compressed. Compression happens afterwards.

If you are going to start throwing in compressed RAW sizes, then you need to compare all of the cameras as compressed RAW values. Which we did do. And by doing so, we do stay under 300MB/s for most cameras. But the values dropped so dramatically and the bandwidth consumed varied so much from camera to camera it brought up the question, "If a camera at half the cost of the A99 can even push 288MB/s compressed, then why is the A99 only pushing half that with its compress RAW files?" This thought, this very thought, brings into question as to whether bandwidth is really the limitation as you claim. But it appears you have ignored this.

Is it really bandwidth that is the limiting factor if the A99 is consuming only around 144 MB/s with its cRAW files at 6fps when the A77 is consuming around 288 MB/s with its cRAW files at 12fps? Explain that one. And again, for simplicities sake, we haven't even touched on the bandwidth consumption of RAW+JPEG.

EinsteinsGhost said:

You've also assumed that in crop mode, the mechanical shutter has a mechanism to drop down to the beginning of "cropped area" and goes no farther than the other end of the cropped area. Got anything to support this claim?

Click to expand...

Now you are trying to change the subject and point it on me. Let's focus on one thing at a time. I gave my theory very early on but we have left that station long ago. You haven't mentioned much about my theory since your first one or two posts to me and even then you had very little to say about it. We have since been focused on your bandwidth limitation theory and you have mentioned very little about my theory during that process. Only JohnFrim has been the one in constant discussion with me about my theory and the flaws in it (hence it being a theory). I look forward to continue my discussion with JohnFrim on my theory and hope to further expand upon it and making corrections along the way.

We have gone very far off the path of comparing theories and have lately only focused on bandwidth limitations. So I do not see the point of going back in and rehashing the other theories out there only to muddy the waters even more than it already is.
--
Paul

VirtualMirage said:

EinsteinsGhost said:

VirtualMirage said:

Let me recap what our discussion has been about and correct me if there is something wrong here:

Click to expand...

Of course. I hate to trim your posts in my responses though but here is why I do it. You start with an assumption, and work with those throughout, no matter how far off they might be from reality. Let us try to work in smaller chunks.

VirtualMirage said:

You were claiming that the frame rate on the A99 was limited by bandwidth. I was telling you that bandwidth is not the limiting factor in the camera. You state that it is and that it appears the maximum throughput of these cameras today caps at around 300MB/s. I then tell you that is not the case and offer some examples where it exceeds that.

Click to expand...

No you didn't. All you did was make assumptions that fit your whim and drew conclusions from it. If Sony A99 is producing 43-44 MB RAW files, and you claim that about 300 MB/s is non-sense, then show me a camera where you see 450-500 MB/s system. Got ONE example?

Click to expand...

I gave you several. You just refuse to accept them. You flip my numbers and began comparing compressed RAW files to uncompressed RAW files so as to suit your 300MB/s figure...

Click to expand...

Don't give me several, give me one that answers my question.

VirtualMirage said:

VirtualMirage said:

You've also assumed that in crop mode, the mechanical shutter has a mechanism to drop down to the beginning of "cropped area" and goes no farther than the other end of the cropped area. Got anything to support this claim?

Click to expand...

Now you are trying to change the subject and point it on me. Let's focus on one thing at a time.

Click to expand...

Sure. But I brought it up because it was your primary argument. We can wait on this one.

EinsteinsGhost said:

Don't give me several, give me one that answers my question.

Click to expand...

Why do I have to repeat myself over and over on the cameras which are listed in the previous comments in this thread that you have already responded to but fail to acknowledge?

The A77 clocks in at 432 MB/s uncompressed at 12fps.

The 1D X clocks in at 376.8 MB/s uncompressed at 12fps. It can go to 14fps but with everything locked up and with the mirror up, without output only in JPEG. But...that JPEG is converted from RAW information. If you want to add what bandwidth is being consumed from the sensor to the processing chip for 14fps then that would bring it to 439.6 MB/s.

the D4 clocks in at 311.3 MB/s uncompressed.

And I failed to ask this in the last post, but why now are you raising your figure from 300MB/s to 450-500MB/s?

I've already shown that cameras can deal with more than 300 MB/s throughput, showing you that the number figure you feel is the ceiling and is effecting the camera's frame rate can already be exceeded in today's cameras. Do you feel you must keep raising the number higher and higher until there are no more cameras to list and then claim you are correct?

If so, that is not the way to prove you are right. That is like changing the rules to a game you keep losing at until all the rules play in your favor and you finally become a victor.

The only thing I can think of that you can do is either prove that about 300 MB/s is the limiting factor (which you have not) or admit that this figure you mentioned was too low and that bandwidth consumption capability is actually higher.

Which if you choose the second option, I will ask again (albeit slightly rephrased):

If bandwidth consumption capability is as high as 450-500 MB/s (the figures you spouted out you wanted me to prove for, which probably includes overhead) and the A99 is only consuming about 252 MB/s at 6fps in uncompressed RAW, then what is limiting its ability to achieve a higher frame rate?

Or

We can even rephrase that question and drop the bandwidth consumption to meet the A77 at 12fps uncompressed to 432 MB/s. Which would look like this:

If bandwidth consumption capability is as high as 432 MB/s and the A99 is only consuming about 252 MB/s at 6fps in uncompressed RAW, then what is limiting its ability to achieve a higher frame rate?

Or

We can even rephrase that question and use compressed RAW bandwidth consumption to meet the A77 at 12fps to 288 MB/s. Which would look like this:

If bandwidth consumption capability is as high as 288 MB/s and the A99 is only consuming about 144 MB/s at 6fps in compressed RAW, then what is limiting its ability to achieve a higher frame rate?
--
Paul

VirtualMirage said:

EinsteinsGhost said:

Don't give me several, give me one that answers my question.

Click to expand...

Why do I have to repeat myself over and over on the cameras which are listed in the previous comments in this thread that you have already responded to but fail to acknowledge?

The A77 clocks in at 432 MB/s uncompressed at 12fps.

Click to expand...

Where did you get the idea that A77 has uncompressed RAW, at 36 MB per file, much less the idea that it applies at 12 fps?

A99 processor. The 'Dual' label just visible refers to card slots not processors.

David

EinsteinsGhost said:

VirtualMirage said:

EinsteinsGhost said:

Don't give me several, give me one that answers my question.

Click to expand...

Why do I have to repeat myself over and over on the cameras which are listed in the previous comments in this thread that you have already responded to but fail to acknowledge?

The A77 clocks in at 432 MB/s uncompressed at 12fps.

Click to expand...

Where did you get the idea that A77 has uncompressed RAW, at 36 MB per file, much less the idea that it applies at 12 fps?

Click to expand...

24 MP (6000x4000) at 12-bits using RGB color gives an uncompressed RAW file of 36 MB, not including any extra variables (lens adjustments, coordinates, etc.) that Sony puts in after the fact.

The A77 will shoot at 12fps in RAW as well as RAW+JPEG in full auto focus mode. The only feature you lose is adjusting the aperture and a bit of a slideshow effect in the EVF. The aperture will lock at wide open or up to F/3.5 when shooting at 12fps. The biggest limitation here is the buffer size and the transfer speed from the buffer to memory card as you are limited to about 11-14 shots (depending on memory card and whether you are shooting RAW or RAW+JPEG) before the full buffer will slow down the frame rate or stop all together.

For comparison, the A900 (which had a slightly higher resolution of 6048 x 4032) had an uncompressed RAW size of 36.6 MB before anything extra and an average file size of about 38.5MB after anything extra is added to it. The A900 was also 12-bit. After compression, the A900 RAW file sizes would be around 25.6 MB, plus or minus. This camera was limited to 5fps with a buffer fill capacity of about 12 RAW images when first released.
--
Paul

It is neither the size of the pixels, nor the amount of data. Though the "amount of data to be handled" is the technical problem.

The actual "why is it slower than" is the cost of making it faster. They easily "could" make it as fast or faster than existing cameras of this sensor / resolution size. They chose not to in order to keep at their price / profit point.

VirtualMirage said:

EinsteinsGhost said:

VirtualMirage said:

EinsteinsGhost said:

Don't give me several, give me one that answers my question.

Click to expand...

Why do I have to repeat myself over and over on the cameras which are listed in the previous comments in this thread that you have already responded to but fail to acknowledge?

The A77 clocks in at 432 MB/s uncompressed at 12fps.

Click to expand...

Where did you get the idea that A77 has uncompressed RAW, at 36 MB per file, much less the idea that it applies at 12 fps?

Click to expand...

24 MP (6000x4000) at 12-bits using RGB color gives an uncompressed RAW file of 36 MB, not including any extra variables (lens adjustments, coordinates, etc.) that Sony puts in after the fact.

Click to expand...

RAW file out of A77 is 24MB. And 24MB at 12 fps = 288 MB/s.

This is exactly the point I was alluding to. You start with the wrong premise to make your point. It is how you ended up with 432 MB/s.

DavieK said:

A99 processor. The 'Dual' label just visible refers to card slots not processors.

David

Click to expand...

Thanks for the shot. Sadly, we can't get any legible information off the chip.

Do we have any other shots that are more in focus or shots of the other side? I wonder how many memory chips they are using?

The A900 used to use two BIONZ chips. I wouldn't be surprised if Sony started using a single chip from the A77 on up that is really a dual or quad core/threaded chip. This would increase the processing power while reduce the amount of hardware that would need to be squeezed onto the circuit board. It should also cut down on power consumption too.
--
Paul

EinsteinsGhost said:

RAW file out of A77 is 24MB. And 24MB at 12 fps = 288 MB/s.

This is exactly the point I was alluding to. You start with the wrong premise to make your point. It is how you ended up with 432 MB/s.

Click to expand...

I didn't start my premise incorrectly. You don't seem to be understanding how RAW files work and how they are handled in-camera. You also failed to elaborate whether or not your 300 MB/s limit was exclusive to just the compressed RAWs or the uncompressed RAWs. I assumed uncompressed since that is the state the RAW file is in prior to reaching the processing chip, thus should be the biggest potential bottleneck when it comes to data throughput.

Here is how you calculate uncompressed RAW file size (using the A77 as the example):

Multiply the horizontal pixels against the vertical pixels:
6,000 pixels x 4,000 pixels = 24,000,000 pixels

Now you multiply the number of pixels by how many bits per pixel:
24,000,000 pixels x 12 bits/pixel = 288,000,000 bits

Now we convert the bits to bytes and then bring the number down to MB:
288,000,000 / 8 bits/byte = 36,000,000 bytes

36,000,000 bytes / 1,000 bytes/KB = 36,000 KB (35,156.25 KB if you divide by 1024)

36,000 KB / 1,000 KB/MB = 36 MB (34.33 MB if you divide 35,156.25 KB by 1024)

That 24MB file size you are mentioning is a compressed RAW file size. What is confusing you is that Sony now only offers compressed RAW as the final output to the memory card. But the data that is being captured from the sensor and being sent to the processing chip is uncompressed RAW (prior to compression) and that data would amount to around 36MB per picture just for the image being captured. After processing is done to the picture it is then compressed and sent to the buffer, which is likely a RAM based memory that is DDR2-800 or faster (which has a theoretical peak throughput of 6400 MB/s). While the pipeline from the processing chip to the buffer probably doesn't handle 6400 MB/s, it is pretty likely that it is faster than 300 MB/s.

If you want to compare compressed RAW file sizes, then the A99 compressed RAW file size is around 24-25MB (at least the file I have looked at was). 24-25MB at 6fps = 144-150 MB/s. So even if you compare both of them in compressed RAW, then there should still be enough bandwidth headroom to account for a faster frame rate assuming its bandwidth capabilities are at least on par to the A77.

Are you sure you are not confusing data bandwidth with processing power?
--
Paul

VirtualMirage said:

EinsteinsGhost said:

RAW file out of A77 is 24MB. And 24MB at 12 fps = 288 MB/s.

This is exactly the point I was alluding to. You start with the wrong premise to make your point. It is how you ended up with 432 MB/s.

Click to expand...

I didn't start my premise incorrectly. You don't seem to be understanding how RAW files work and how they are handled in-camera. You also failed to elaborate whether or not your 300 MB/s limit was exclusive to just the compressed RAWs or the uncompressed RAWs. I assumed uncompressed since that is the state the RAW file is in prior to reaching the processing chip, thus should be the biggest potential bottleneck when it comes to data throughput.

Here is how you calculate uncompressed RAW file size (using the A77 as the example):

Multiply the horizontal pixels against the vertical pixels:
6,000 pixels x 4,000 pixels = 24,000,000 pixels

Now you multiply the number of pixels by how many bits per pixel:
24,000,000 pixels x 12 bits/pixel = 288,000,000 bits

Now we convert the bits to bytes and then bring the number down to MB:
288,000,000 / 8 bits/byte = 36,000,000 bytes

36,000,000 bytes / 1,000 bytes/KB = 36,000 KB (35,156.25 KB if you divide by 1024)

36,000 KB / 1,000 KB/MB = 36 MB (34.33 MB if you divide 35,156.25 KB by 1024)

That 24MB file size you are mentioning is a compressed RAW file size. What is confusing you is that Sony now only offers compressed RAW as the final output to the memory card. But the data that is being captured from the sensor and being sent to the processing chip is uncompressed RAW (prior to compression) and that data would amount to around 36MB per picture just for the image being captured. After processing is done to the picture it is then compressed and sent to the buffer, which is likely a RAM based memory that is DDR2-800 or faster (which has a theoretical peak throughput of 6400 MB/s). While the pipeline from the processing chip to the buffer probably doesn't handle 6400 MB/s, it is pretty likely that it is faster than 300 MB/s.

If you want to compare compressed RAW file sizes, then the A99 compressed RAW file size is around 24-25MB (at least the file I have looked at was). 24-25MB at 6fps = 144-150 MB/s. So even if you compare both of them in compressed RAW, then there should still be enough bandwidth headroom to account for a faster frame rate assuming its bandwidth capabilities are at least on par to the A77.

Are you sure you are not confusing data bandwidth with processing power?
--
Paul

Click to expand...

You don't need to do the math, you simply need to take a look at what A77 writes out: a 24MB RAW file (not 36 MB). That is because A77 produces cRAW (not pure RAW), which produces smaller file size. This does not apply to A99, which not only has 14-bit RAW, but it ain't cRAW.

EinsteinsGhost said:

You don't need to do the math, you simply need to take a look at what A77 writes out: a 24MB RAW file (not 36 MB). That is because A77 produces cRAW (not pure RAW), which produces smaller file size.

Click to expand...

If you are not doing the math, then you must have failed a lot of math classes. j/k

EinsteinsGhost said:

This does not apply to A99, which not only has 14-bit RAW, but it ain't cRAW.

Click to expand...

How do you know? Have you looked at the A99 RAW files?

I have looked at the A99 RAW files, and the ones I have viewed were around 24-25MB in size. This is clearly an indication that the A99 RAW files are in a cRAW format as well. Remember, the A77 doesn't label their RAW files as compressed, but they clearly are due to the files sizes they output at.

One of the great things about RAW files in their early days (and still to this day) is that the data was not compressed. It was exactly how the data was captured from the sensor and converted by the processor. RAW files captured every bit for every pixel, used or not. This was the downside to RAW too since files sizes can be much larger, no matter what image was taken. You could take a picture of of a white backdrop, the inside of your lens cap, or the landscape of Sicily and the files sizes would vary very little.

But as file sizes got larger due to higher resolution sensors, camera manufacturers began finding ways to compress the RAW file after processing to lower its size through multiple methods. Some would convert the 12-bit to a 10-bit with an associated lookup table to fill in the blanks, others would eliminate the unneeded data at the peaks and use a simple bit for this information, and so on. This gave them the ability to compress the file into a lossless/lossy format. The concerns by doing this is what was the toll that was paid in the ability to recover in certain extremes and did it effect the gradient capabilities with certain scenes? Some say yes it does, others say it has no effect. This is still up for debate.

The greater problem now is that camera manufacturers are now not giving us a choice to choose. This is what they did with the A77. The sensor is capturing full sized RAW information and handing it over to the processing chip. So the information is there, but they now force a form of cRAW on us with no choice.

So, yes, it does apply to the A99. How that data is being handled on the A77 versus the A99 is no different.
--
Paul